This invention relates to electronic radio frequency identification (RFID) systems and more particularly to such systems including at least one interrogator or reader and a plurality of transponders.
In known systems of the aforementioned kind, the interrogator is caused to transmit a radio frequency (RF) interrogation signal towards the transponders. The transponders respond to this signal by backscatter modulation of the interrogation signal with a data message including base data which may include an identification code characteristic of the transponder.
During transmission by the transponders a brief burst of high-intensity radio noise from any source may temporarily overload the radio receiver of the interrogator. A noise burst may result from sources including natural phenomena, other active transponders in the interrogator antenna field sending data simultaneously, or signals emitted by other radio services not associated with the RFID system. The effect is to corrupt the data message received for at least some of the duration of the noise burst, resulting in a “burst error” in the message. In a digital network, a noise burst typically results in a burst error in which some or all of a consecutive sequence of data bits are received incorrectly.
A second kind of burst error may occur in the transmitted data bits when the signal is temporarily below the threshold of the receiver of the interrogator due to propagation fading. Alternatively, signal drop out may occur due to cancellation caused by multi-path transmission between the transponder and the interrogator. The net effect on the transmitted data is the same in that a burst error occurs where some or all of a number of consecutive base data bits are received incorrectly.
It has been found that typical burst errors in transmissions of the kind utilized by the system and method according to the invention are four (4) to six (6) bits in length. Since burst errors are a statistical phenomena, this means that there will be many which are shorter and some which may be much longer than 4-6 bits. The burst error length and the average interval between bursts vary with frequency and noise origin. Therefore, any method utilized for correction or compensation for burst errors ideally must be robust enough to handle one or multiple burst errors of varying length within a single data message.
Several approaches may be taken to deal with the effect of burst errors in a digital data message. The first step is to detect that they have occurred. Digital error-detection techniques such as parity checking and the CCITT Cyclic Redundancy Check (CRC) have been used for error detection in systems that use digital messages to carry the base data. A common approach to ensuring accurate communication is to use error detection in conjunction with on-demand retransmission (such as an ACK/NAK protocol) as a method of ensuring accurate data message receipt. The message is re-transmitted until no error is detected. This method works as long as there are some clear, noise-free intervals long enough for the entire message to fit into and to be received without error—when the interval between noise bursts is less than the message length, it cannot succeed. For this reason, long messages are usually broken into a sequence of many short data packets, each of which is treated as a separate ordered component of the entire method and transmitted separately.
Once all the packets have been received correctly, the complete base data can be accurately reassembled.
A major weakness of this method is that increased communications time is required, both from the extra communications overhead of packet transmission and from the number of retransmissions which may be required for the entire single message (or all its component packets) to be received once without error. A second weakness of this method is that it requires that the data message source be in two-way communication with the receiver. This is not always feasible, as in deep space communications.
Forward Error Correction (FEC) is used when data must be sent over noisy channels, and under situations such as deep space communications, where on-demand retransmission is not feasible. This involves sending some redundant information along with the required base data as part of the entire data message. The receiver can determine from the totality of data sets received in the message whether the base data was correctly received, and use these additional data sets to detect and correct erroneous data bits until the base data can be reconstructed without error, or with an acceptable level of accuracy.
One approach to FEC is to multiply retransmit the base data within the data message, possibly as algorithmically transformed, and apply voting techniques to determine the correct base data bits in the presence of noise. This method was used by Lowe in U.S. Pat. No. 5,742,618. The main disadvantage of this is that it requires multiple copies of the base data to be encoded in the data message as a means of ensuring accurate reception, greatly increasing the transmission time. In addition, while this method requires little computational capability it is non-deterministic in error correction. It requires only that a majority of the received data sets agree; regardless of whether or not they were correctly received.